To the Editor.
McKenna, Mosko, and Richard compare nocturnal breastfeeding patterns of 35 infants while both sleeping with their mothers and while sleeping alone in a laboratory setting.1 Although the results of the study are interesting and it is difficult to argue with the desirability of enhancing breastfeeding in relation to a number of infant outcomes, the authors draw conclusions about the possible protective effect of bedsharing against sudden infant death syndrome (SIDS) that have no foundation, based either on their own study or the work of numerous other researchers.
The literature on breastfeeding and SIDS is inconsistent. Gilbert and coauthors summarized the results of 17 case-control and 1 cohort study that analyzed feeding methods in relation to SIDS.2 Eleven showed an increased risk of SIDS in bottle-fed babies and 7 showed no relationship. Inconsistent findings were related to different ways of measuring feeding and different ways of considering potential confounding factors, such as socioeconomic variables. Gilbert et al found in their own study of SIDS in England that breastfeeding was not protective after taking into account maternal smoking, parental employment, preterm gestation, and sleep position. In Chicago, where a case-control study of SIDS risk factors has recently concluded, preliminary analysis of the relationship between never breastfeeding and SIDS, based on 250 SIDS cases and 250 matched living controls, showed an odds ratio of 2.8 (95% confidence interval, 1.7–4.6), adjusting for socioeconomic status.4 However, after adjusting for additional factors, such as maternal smoking, bed softness, and sleep position, it was no longer significant.
Rates of SIDS have declined dramatically in a number of countries, including England, New Zealand, and Australia, following education campaigns focusing on sleep position and other recommendations.5 The single most important factor accounting for the declining rates was reduced prevalence of prone sleep position.5 Rates of breastfeeding did not change significantly. For example, in Tasmania, Australia, 70% of the reduction in SIDS rate could be accounted for by the decreased prone position prevalence.6 Rates of bottle-feeding at 1 month remained unchanged at 66%. In the United States, declining rates of SIDS have also been attributed in large part to declining use of the prone position, following recommendations made by the American Academy of Pediatrics (AAP) and the Back to Sleep Campaign.7,,8Therefore, based on these studies and international trends in SIDS rates, there is insufficient evidence to support that breastfeeding per se reduces the risk of SIDS.
In the same issue of Pediatrics is a statement from the AAP Task Force on Infant Positioning and SIDS related to bedsharing and SIDS.9 As pointed out by the Task Force, no studies have demonstrated a protective effect of bedsharing, and others have shown an increased risk, at least under certain circumstances. For example, the New Zealand Cot Death Study found that bedsharing was a risk factor for SIDS among infants whose mothers smoked. As presented at the Infant Sleep Environment and SIDS Risk Conference sponsored by the National Institute of Child Health and Human Development, the Chicago study, one of the few large epidemiologic studies examining current SIDS risk factors in the United States, found in preliminary analyses that bedsharing with one or both parents on the night of the infant's death presented a twofold increased risk for SIDS, after adjusting for socioeconomic status.10 Bedsharing with siblings (with or without the parents) raised the risk fivefold. Bedsharing was present among 51% of the infants who died from SIDS and 30% of controls during a comparable designated sleep (P < .05). In St Louis, bedsharing among African-American families whose infants are at higher risk for SIDS has been found to be as high as 60%.11
It is also essential to consider the circumstances under which bedsharing is occurring. In McKenna and coauthors' study, the environment and selection of subjects were highly controlled. Fathers, for example, were excluded. We do not know if the results would be influenced by the presence of the father or other family members who may routinely bedshare with the infant at home. Furthermore, their subjects were infants at lower risk for SIDS, ie, normal birth weight, full gestation, and normal growth and development. Drawing conclusions about bedsharing and SIDS risk based on this highly selected population is inappropriate.
Finally, McKenna and his colleagues argue that the marginally lower rates of SIDS among Latino infants compared with white infants may be related to increased bedsharing more common to Latino mothers and infants. Again, this is pure speculation and could be related to a number of other factors, such as lower smoking rates among Latinos.
Careful analysis of epidemiologic data, in which adjustment for multiple related factors is possible, is essential to further our understanding of the complicated causative pathway that leads to SIDS. Unfortunately, the research presented by McKenna and coauthors, though perhaps important in terms of increasing breastfeeding, does not contribute to our understanding of SIDS risk or provide valid insights about sleep practices that might reduce this risk.
- McKenna JJ,
- Mosko SS,
- Richard CA
- Gilbert RE,
- Wigfield RE,
- Fleming PJ,
- Berry PJ,
- Rudd PT
- ↵Hauck FR, Merrick CA, Donovan JM, Iyasu S, Willinger M. Risk factors for sudden infant death syndrome: preliminary results from the Chicago Infant Mortality Study (Abstract). American Public Health Association 124th Annual Meeting. November 17–21, 1996; New York, NY
- ↵Bak SM, Willinger M, Hoffman HJ, et al. Infant sleep practices: results from US national surveys 1992–1995 (Abstract). Fourth SIDS International Conference; June 23–26, 1996; Bethesda, MD
- American Academy of Pediatrics Task Force on Infant Positioning and SIDS
- ↵Hauck FR. Bedsharing: review of epidemiologic data examining links to SIDS (Abstract). Infant Sleep Environment and SIDS Risk Conference. January 9–10, 1997; Bethesda, MD
- ↵Kemp JS. The bedsharing experience (Abstract). Infant Sleep Environment and SIDS Risk Conference. January 9–10, 1997; Bethesda, MD
- Copyright © 1998 American Academy of Pediatrics
I couldn't agree more with Drs Hauck and Kemp's words of caution that there is no evidence that bedsharing reduces the risk of SIDS and that under certain circumstances bedsharing may actually increase the risk of infant death. They are also correct that the previously demonstrated protective effect of breastfeeding on SIDS weakens considerably after controlling for confounding variables. I am disappointed that McKenna et al expressed any implications about SIDS in their report, because although breastfeeding was increased when mothers slept with their infants according to the study conditions, the authors presented no evidence of any relationship between SIDS and bedsharing. There are many other legitimate reasons why encouraging breastfeeding is a good thing. Currently, the most effective means of reducing SIDS risk is to place infants down to sleep on their backs.
Our interest in whether breastfeeding in the bedsharing environment might reduce vulnerability to SIDS stems in part from the unexplained observations that SIDS rates can be much lower in societies or cultural groups that practice forms of parent-infant cosleeping often with breastfeeding such as in Hong Kong,1,,2Japan,3 and areas of London in which recent Bangladesh immigrants continue the custom of cosleeping.4–6Moreover, that close proximity to parents during sleep contributes to these low SIDS rates in these countries is suggested by the New Zealand study cited by Hauck and Kemp in their critique. Infants who slept in a room alone were nearly four times as likely to die from SIDS as infants who shared a room with adult(s), but this protective effect did not generalize to room-sharing with siblings.7 This suggests that an active role by an adult caregiver is necessary before cosleeping (and/or bedsharing) can be beneficial, a finding consistent with our hypothesis.8,,9
Furthermore, the recent increase in SIDS rates in Japan, which has been paralleled by a shift from a tradition of social sleeping to solitary sleeping, also supports a potentially protective role of parental proximity during sleep.10 Amongst the Pacific Islanders, who commonly bedshare, SIDS is lower than any other group in New Zealand, further suggesting that at least under some circumstances proximity to the parent(s) during sleep may be protective, as we are proposing.7
The epidemiologic data provide evidence of the adverse effects of smoking and other risk factors that can accompany bedsharing, but risks are not increased when maternal smoking is absent,11,,12and no research to date demonstrates that SIDS risks are increased among nonsmoking mothers who breastfeed and share their beds with their babies.
It is precisely an appreciation of how particular combinations of factors combine to effect outcomes relating to bedsharing, and how difficult they are to completely decompose statistically, that lead us to argue that global, unqualified statements claiming that bedsharing increases SIDS risks are inaccurate and both scientifically and morally unjustified.13,,14Breastfeeding while bedsharing, as compared with bedsharing without breastfeeding, as compared with bedsharing without breastfeeding in combination with smoking, represent only three of many distinct “types” of bedsharing. Each type needs to be distinguished and studied in relationship to mattress firmness and infant sleep position (to name only few factors) if epidemiologic studies of bedsharing are to be useful. Because specific conditions define where bedsharing can be dangerous, is it not then possible that specific conditions can define where bedsharing can be protective?
Knowing something about each population and how bedsharing is expressed is critical in assessing whether or not the particular risks elucidated are necessarily relevant to other groups. The Chicago Infant Mortality Study, where Hauck found a twofold increased risk associated with bedsharing, is a case in point. Both SIDS victims' families and controls are members of an urban economic underclass that suffer at every level from high age-specific mortality and poorer overall health than most, if not all, other groups. A recent study of demographic and mortality data for 77 community areas in Chicago encompassed by Hauck's sample found a correlation of −88 between average life expectancy ranging from 54.3 to 77.4 years and neighborhood homicide rate. Furthermore, those neighborhoods with the shortest average life expectancy also showed the highest birth rates for young-age minorities relative to areas with the largest life expectancies.15 Relatedly, the families of SIDS victims and controls include people attempting to subsist on incomes lower than the poverty level and, in some case, on incomes as low as $2000 per year.16 Clearly, the meaning of “controlling” for factor X or Y in this socially and economically disadvantaged population means something quite different from a population characterized by more normally distributed incomes based on national averages. Breastfeeding practiced by nonsmoking/bedsharing mothers, in this population, is practically nonexistent. Therefore, it is hardly a population from which reliable inferences about the potentially protective functions of bedsharing with breastfeeding in nonsmoking environments can be drawn.
That Hauck's bedsharing analysis failed to distinguish between bedsharing in a bed and mothers and infants sleeping together on a sofa16 is even more problematic and renders the use of her findings inappropriate for evaluating our hypothesis. The vertical side of a sofa against which infants can be pinned during sleep can cause the infant to slip face down into a wedge at the back of the cushion and suffocate. Because suffocation and SIDS are hard to distinguish, such a death could easily be a proxy for a “bedsharing SIDS” when in fact it was more likely suffocation. This definition of bedsharing illustrates even more why simple generalizations or conclusions about the supposed universal risks of “bedsharing” must be limited to particular circumstances and/or populations, at least for now. Definitions vary widely across studies, including several that define bedsharing as “bedsharing with anyone,” and not necessarily with a responsible adult. Surely, who the infant is sleeping with likely affects outcome.11
As I argued at the AAP Task Force on Infant Positioning and SIDS (on bedsharing) the variable “bedsharing” is too gross a descriptor to have any direct function or outcome assigned to it, because to be useful, definitions must capture what actually happens (or fails to happen) to an infant once in bed. Bedsharing is not binary or a categorical equivalent of an infant sleeping prone or supine, although it is treated epidemiologically as if it were.17
Hauck and Kemp suggest that our hypothesis lacks foundation. As we acknowledged twice in our article, increased protection from SIDS through breastfeeding is not universally established but at least half the studies show it as being protective, and because no two studies use the same definition of breastfeeding, our proposal is justified. Only one epidemiologic study has observed whether such dose-specific response effects exist and whether they are stable across races and socioeconomic groups in relationship to SIDS. The results of the analysis strengthen the foundation for our proposal. Using the National Center for Health Statistics National Maternal and Infant Health Survey, Fredrickson et al18 analyzed 7102 controls, 499 SIDS, and 584 non-SIDS deaths. They defined breastfeeding categorically (with various cutoffs) and as a continuous variable (months to final weaning and exclusive breastfeeding month equivalents—the only epidemiologic study to do so. Univariate odds ratio and logistic regression to control for a variety confounders were used. For both black and white Americans the risk of SIDS decreased for every month of breastfeeding. Conversely, for white mothers the risk of SIDS increased by 1.19 for every month of not breastfeeding, and 2.0 for every month of nonexclusive breastfeeding. For black mothers, the risk of SIDS also increased by 1.19 for every month of not breastfeeding, but 2.3 for every month of not exclusively breastfeeding.18
These results suggest that simple epidemiologic questions such as “never breastfed” used by Hauck in her study cannot answer the question of whether or not breastfeeding is ever protective. Absence of proof (especially in Hauck's unique population) is not proof of its absence.
Hauck and Kemp rightly stress how critical the change to the supine infant sleep position has been in reducing SIDS rates worldwide. Our data show convincingly that to facilitate access to the breast compared with routinely solitary sleeping infants, the safer supine infant sleep position is always expressed, and without instruction, among routinelybreastfeeding bedsharing mothers and infants, further strengthening our hypothesis.19
If the AAP statement on bedsharing to which Hauck and Kemp refer leaves the impression that all bedsharing functions the same way or that mere physical proximity to the mother for sleep and breastfeeding, on the same surface, is inherently dangerous, I submit that it is at the very least imprecise and at its worst misleading. The AAP statement is not meant to imply that the fundamental biological linkages between mothers and infants sleeping side by side to breastfeed are intrinsically and inevitably lethal. It says only that controlled epidemiologic studies have not yet shown that bedsharing (without breastfeeding) is protective, and that particular adverse conditions within which bedsharing can occur can increase SIDS risk, just as we have argued many times elsewhere.8,,9,13,17
- Lee NP,
- Chann YF,
- Davis DP,
- Lau E,
- Yip DCP
- Tasaki H,
- Yamashita M,
- Miyazaki S
- Gantly M,
- Davies DP,
- Murcott A
- Farooqui S,
- Perry LJ
- Balararian R,
- Raleigh VS,
- Botting B
- Mosko S,
- Richard C,
- McKenna J
- Kloneff-Cohen H,
- Edelstein S
- Fleming P,
- Blair P,
- Bacon C
- 2-13.↵McKenna JJ. The potential benefits of infant-parent cosleep in relation to SIDS prevention: overview and critique of epidemiological bed sharing studies. In: Rognum TO, ed. Sudden Infant Death Syndrome: New Trends in the 90's. Oslo, Norway: Scandinavia University Press; 1995
- McKenna JJ,
- Mosko S,
- Richard C
- Wilson M,
- Daly M
- 2-16.↵Hauck F. Bedsharing: review of epidemiological data examining links to SIDS (Abstract). Infant Sleep Environment and SIDS Risks Conference. January 9–10, 1997; Bethesda, MD
- 2-17.↵McKenna JJ. Sleeping arrangements defined and the bedsharing experience: relevance to SIDS (Abstract). Infant Sleep Environment and SIDS Risk Conference. January 9–10, 1997; Bethesda, MD
- 2-18.↵Fredrickson DD, Sorenson JR, Biddle AK, Kotelchack M. Relationship of sudden infant death syndrome of breast-feeding duration and intensity. Am J Dis Child. 1993;147:460. Abstract No. 191
- Copyright © 1998 American Academy of Pediatrics